Open dataset v1
The Case Study 01 relational dataset: 52 decisions, 37 matters, pseudonymized, with a data dictionary, controlled vocabularies and a confidence score on every extracted field. A silent field is recorded as silent, never as zero. Release accompanies the case study; the dataset page will host the download.
Known undercount: the OPP layer
First, what the third corpus is. Our third case study is Ontario corrections — 484 decisions, 344 matters, 1984–2026, with the Crown in right of Ontario as the single respondent throughout. It is a study of a provincial ministry, not of a police service, and it never appears on any police per-capita measure. Facility names in it — Toronto East, Toronto South, Toronto Jail — are provincial institutions that happen to sit in a city; the city name is a location, not a municipal body.
But it exposed a gap in the police record. Of those 484 decisions, 80 — 17% — carry a police signal: matters involving policing that are captioned against the corrections ministry, with respondent strings containing no policing word at all. These are predominantly OPP matters ◇ — the exact OPP share within the 80 is being counted and is not yet published.
Why name-matching fails here. The respondent is a single ministry that has been renamed five times. From 2003 to April 3, 2019 it was the Ministry of Community Safety and Correctional Services; since April 4, 2019 it has been the Ministry of the Solicitor General; earlier names include the Ministry of Correctional Services and the Ministry of Public Safety and Security. Not one of those names contains a policing word — yet the ministry is directly responsible for the Ontario Provincial Police. So an OPP matter is correctly captioned to a ministry whose name gives no indication that policing is involved, and any extraction that matched on respondent name, including ours, missed every one of them.
Two of the 80 are remedy-bearing: Prothero, 2016 HRTO 1481 (a $25,000 award against the OPP as employer, plus $2,500 payable personally by a named individual and an order that he complete Commission training) and Kurpel, 2021 HRTO 83. Others include an OPP death-threat investigation, an OPP employment and PTSD matter, and an OPP attendance at a property — all captioned to a ministry rather than to a police service.
The corpus-gap registry
What each source can see: CanLII publishes decisions but prohibits automated collection and lists no file numbers on its indexes. A2AJ, Canada's leading open-legal-data project, holds the full federal CHRT corpus — and no HRTO, and no Ontario Divisional Court. Tribunals Ontario posts quarterly volume CSVs at predictable URLs that are not listed in the Ontario Data Catalogue. Applications that never produce a decision appear in no public source at all. Canada's open-legal-data infrastructure cannot see Ontario's human rights record — that is a finding, not a footnote.
The ledgers and the gap
What changed / where it went
Eighteen years of narrowings, beside the money spent resisting accountability. Two phones.
R·35The public-interest gap
The claim nobody is allowed to bring — and our formal request that the Commission bring it.
R·PRWhich cases actually get cited
Gate authorities beat substantive ones 5:1. Intersectionality: zero of 484 decisions.
R·3DThree doors
Profiling law is built by accused people excluding evidence — not by complainants. What that means for forum choice.
R·STWhat the statute actually says
We read TARA instead of guessing. A register of proceedings is expressly a public record.
R·WDWho decides
Appointments, independence, the 2018 complement collapse, and concentration reported without names.
R·97Nearly 97% dismissed without a hearing
The province-wide figures, Bokhari, the 2019 consolidation, and the cases in the news.
R·2SOne claim, two systems
The federal tribunal decided the claim Ontario refused to hear — and the two regimes that fund policing.
R·NRNamed, then removed
Where the individual respondents go. Eleven removed from one cluster; zero personal findings in eighteen years.